From ca456ae280c0646e1e571c3b9a3834c55e90adfe Mon Sep 17 00:00:00 2001 From: liubo Date: Wed, 1 Jun 2011 09:42:49 +0000 Subject: Btrfs: don't save the inode cache in non-FS roots This adds extra checks to make sure the inode map we are caching really belongs to a FS root instead of a special relocation tree. It prevents crashes during balancing operations. Signed-off-by: Liu Bo Signed-off-by: Chris Mason --- fs/btrfs/inode-map.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'fs/btrfs/inode-map.c') diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 3262cd17a12f..04f7199facb4 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -388,6 +388,12 @@ int btrfs_save_ino_cache(struct btrfs_root *root, int prealloc; bool retry = false; + /* only fs tree and subvol/snap needs ino cache */ + if (root->root_key.objectid != BTRFS_FS_TREE_OBJECTID && + (root->root_key.objectid < BTRFS_FIRST_FREE_OBJECTID || + root->root_key.objectid > BTRFS_LAST_FREE_OBJECTID)) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; -- cgit v1.2.3 From d132a538d258f8f52fd0cd8b5017755f4e915386 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 31 May 2011 19:33:33 +0000 Subject: Btrfs: don't save the inode cache if we are deleting this root With xfstest 254 I can panic the box every time with the inode number caching stuff on. This is because we clean the inodes out when we delete the subvolume, but then we write out the inode cache which adds an inode to the subvolume inode tree, and then when it gets evicted again the root gets added back on the dead roots list and is deleted again, so we have a double free. To stop this from happening just return 0 if refs is 0 (and we're not the tree root since tree root always has refs of 0). With this fix 254 no longer panics. Thanks, Signed-off-by: Josef Bacik Tested-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/inode-map.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'fs/btrfs/inode-map.c') diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 04f7199facb4..2d0d50067a7b 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -394,6 +394,11 @@ int btrfs_save_ino_cache(struct btrfs_root *root, root->root_key.objectid > BTRFS_LAST_FREE_OBJECTID)) return 0; + /* Don't save inode cache if we are deleting this root */ + if (btrfs_root_refs(&root->root_item) == 0 && + root != root->fs_info->tree_root) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; -- cgit v1.2.3 From 4b9465cb9e3859186eefa1ca3b990a5849386320 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 3 Jun 2011 09:36:29 -0400 Subject: Btrfs: add mount -o inode_cache This makes the inode map cache default to off until we fix the overflow problem when the free space crcs don't fit inside a single page. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/free-space-cache.c | 6 ++++++ fs/btrfs/inode-map.c | 20 ++++++++++++++++++++ fs/btrfs/super.c | 8 +++++++- 4 files changed, 34 insertions(+), 1 deletion(-) (limited to 'fs/btrfs/inode-map.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8f98c2005715..4958ef5417d6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1340,6 +1340,7 @@ struct btrfs_ioctl_defrag_range_args { #define BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED (1 << 14) #define BTRFS_MOUNT_ENOSPC_DEBUG (1 << 15) #define BTRFS_MOUNT_AUTO_DEFRAG (1 << 16) +#define BTRFS_MOUNT_INODE_MAP_CACHE (1 << 17) #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 1cb72394498c..bffa5c4a633b 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -2536,6 +2536,9 @@ int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root) int ret = 0; u64 root_gen = btrfs_root_generation(&root->root_item); + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + /* * If we're unmounting then just return, since this does a search on the * normal root and not the commit root and we could deadlock. @@ -2575,6 +2578,9 @@ int btrfs_write_out_ino_cache(struct btrfs_root *root, struct inode *inode; int ret; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + inode = lookup_free_ino_inode(root, path); if (IS_ERR(inode)) return 0; diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 2d0d50067a7b..cb79b8975c9f 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -38,6 +38,9 @@ static int caching_kthread(void *data) int slot; int ret; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; @@ -141,6 +144,9 @@ static void start_caching(struct btrfs_root *root) int ret; u64 objectid; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return; + spin_lock(&root->cache_lock); if (root->cached != BTRFS_CACHE_NO) { spin_unlock(&root->cache_lock); @@ -178,6 +184,9 @@ static void start_caching(struct btrfs_root *root) int btrfs_find_free_ino(struct btrfs_root *root, u64 *objectid) { + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return btrfs_find_free_objectid(root, objectid); + again: *objectid = btrfs_find_ino_for_alloc(root); @@ -201,6 +210,10 @@ void btrfs_return_ino(struct btrfs_root *root, u64 objectid) { struct btrfs_free_space_ctl *ctl = root->free_ino_ctl; struct btrfs_free_space_ctl *pinned = root->free_ino_pinned; + + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return; + again: if (root->cached == BTRFS_CACHE_FINISHED) { __btrfs_add_free_space(ctl, objectid, 1); @@ -250,6 +263,9 @@ void btrfs_unpin_free_ino(struct btrfs_root *root) struct rb_node *n; u64 count; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return; + while (1) { n = rb_first(rbroot); if (!n) @@ -399,9 +415,13 @@ int btrfs_save_ino_cache(struct btrfs_root *root, root != root->fs_info->tree_root) return 0; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; + again: inode = lookup_free_ino_inode(root, path); if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) { diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 28e3cb2607ff..3559d0b3518a 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -160,7 +160,8 @@ enum { Opt_compress_type, Opt_compress_force, Opt_compress_force_type, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, - Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_err, + Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, + Opt_inode_cache, Opt_err, }; static match_table_t tokens = { @@ -192,6 +193,7 @@ static match_table_t tokens = { {Opt_enospc_debug, "enospc_debug"}, {Opt_subvolrootid, "subvolrootid=%d"}, {Opt_defrag, "autodefrag"}, + {Opt_inode_cache, "inode_cache"}, {Opt_err, NULL}, }; @@ -360,6 +362,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) printk(KERN_INFO "btrfs: enabling disk space caching\n"); btrfs_set_opt(info->mount_opt, SPACE_CACHE); break; + case Opt_inode_cache: + printk(KERN_INFO "btrfs: enabling inode map caching\n"); + btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); + break; case Opt_clear_cache: printk(KERN_INFO "btrfs: force clearing of disk cache\n"); btrfs_set_opt(info->mount_opt, CLEAR_CACHE); -- cgit v1.2.3 From 7841cb2898f66a73062c64d0ef5733dde7279e46 Mon Sep 17 00:00:00 2001 From: David Sterba Date: Tue, 31 May 2011 18:07:27 +0200 Subject: btrfs: add helper for fs_info->closing wrap checking of filesystem 'closing' flag and fix a few missing memory barriers. Signed-off-by: David Sterba --- fs/btrfs/ctree.h | 9 +++++++++ fs/btrfs/extent-tree.c | 3 +-- fs/btrfs/file.c | 4 ++-- fs/btrfs/free-space-cache.c | 10 ++++------ fs/btrfs/inode-map.c | 3 +-- fs/btrfs/inode.c | 3 +-- fs/btrfs/scrub.c | 2 +- fs/btrfs/transaction.c | 2 +- 8 files changed, 20 insertions(+), 16 deletions(-) (limited to 'fs/btrfs/inode-map.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 4958ef5417d6..8490ee063709 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -2354,6 +2354,15 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *node, struct extent_buffer *parent); +static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info) +{ + /* + * Get synced with close_ctree() + */ + smp_mb(); + return fs_info->closing; +} + /* root-item.c */ int btrfs_find_root_ref(struct btrfs_root *tree_root, struct btrfs_path *path, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c9173a7827b0..5b9b6b6df242 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -366,8 +366,7 @@ again: nritems = btrfs_header_nritems(leaf); while (1) { - smp_mb(); - if (fs_info->closing > 1) { + if (btrfs_fs_closing(fs_info) > 1) { last = (u64)-1; break; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 982b5ea9762f..fa4ef18b66b1 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -129,7 +129,7 @@ int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans, if (!btrfs_test_opt(root, AUTO_DEFRAG)) return 0; - if (root->fs_info->closing) + if (btrfs_fs_closing(root->fs_info)) return 0; if (BTRFS_I(inode)->in_defrag) @@ -229,7 +229,7 @@ int btrfs_run_defrag_inodes(struct btrfs_fs_info *fs_info) first_ino = defrag->ino + 1; rb_erase(&defrag->rb_node, &fs_info->defrag_inodes); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) goto next_free; spin_unlock(&fs_info->defrag_inodes_lock); diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index bffa5c4a633b..ad144736a5fd 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -98,7 +98,7 @@ struct inode *lookup_free_space_inode(struct btrfs_root *root, return inode; spin_lock(&block_group->lock); - if (!root->fs_info->closing) { + if (!btrfs_fs_closing(root->fs_info)) { block_group->inode = igrab(inode); block_group->iref = 1; } @@ -493,8 +493,7 @@ int load_free_space_cache(struct btrfs_fs_info *fs_info, * If we're unmounting then just return, since this does a search on the * normal root and not the commit root and we could deadlock. */ - smp_mb(); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) return 0; /* @@ -2513,7 +2512,7 @@ struct inode *lookup_free_ino_inode(struct btrfs_root *root, return inode; spin_lock(&root->cache_lock); - if (!root->fs_info->closing) + if (!btrfs_fs_closing(root->fs_info)) root->cache_inode = igrab(inode); spin_unlock(&root->cache_lock); @@ -2543,8 +2542,7 @@ int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root) * If we're unmounting then just return, since this does a search on the * normal root and not the commit root and we could deadlock. */ - smp_mb(); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) return 0; path = btrfs_alloc_path(); diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index cb79b8975c9f..b4087e0fa871 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -62,8 +62,7 @@ again: goto out; while (1) { - smp_mb(); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) goto out; leaf = path->nodes[0]; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index a83e44bf3206..02ff4a1b968b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4266,8 +4266,7 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc) if (BTRFS_I(inode)->dummy_inode) return 0; - smp_mb(); - if (root->fs_info->closing && is_free_space_inode(root, inode)) + if (btrfs_fs_closing(root->fs_info) && is_free_space_inode(root, inode)) nolock = true; if (wbc->sync_mode == WB_SYNC_ALL) { diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c index 1204eab94028..df50fd1eca8f 100644 --- a/fs/btrfs/scrub.c +++ b/fs/btrfs/scrub.c @@ -1183,7 +1183,7 @@ int btrfs_scrub_dev(struct btrfs_root *root, u64 devid, u64 start, u64 end, int ret; struct btrfs_device *dev; - if (root->fs_info->closing) + if (btrfs_fs_closing(root->fs_info)) return -EINVAL; /* diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 2d5c6d2aa4e4..dd719662340e 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -817,7 +817,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) btrfs_btree_balance_dirty(info->tree_root, nr); cond_resched(); - if (root->fs_info->closing || ret != -EAGAIN) + if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN) break; } root->defrag_running = 0; -- cgit v1.2.3